package rtcconformance import ( "bytes" "crypto/sha256" "encoding/hex" "encoding/json" "errors" "fmt" "io" "regexp" "slices" "strings" "time" ) const ( CaptureSchema = "real-device-rtc-capture-v1" ReportSchema = "real-device-rtc-conformance-report-v1" MinimumDevices = 3 MinimumSessions = 6 ) var digestPattern = regexp.MustCompile(`^sha256:[0-9a-f]{64}$`) var RequiredEngineModes = []string{"s2s_pure", "s2s_hybrid", "cascade"} var RequiredScenarios = []string{"NORMAL_COMPLETION", "TOKEN_RENEWAL", "NETWORK_SWITCH_RECOVERY", "CONTROLLED_STOP"} var RequiredLifecycle = []string{"ROOM_ISSUED", "STARTING_AGENT", "ACTIVE", "STOPPING", "COMPLETED"} type Capture struct { SchemaVersion string `json:"schema_version"` EvidenceSource string `json:"evidence_source"` CandidateVersion string `json:"candidate_version"` BuildDigest string `json:"build_digest"` Environment string `json:"environment"` RTCProviderRef string `json:"rtc_provider_ref"` VoiceProviderRef string `json:"voice_provider_ref"` RTCApplicationRef string `json:"rtc_application_ref"` RegionRef string `json:"region_ref"` ChangeTicketRef string `json:"change_ticket_ref"` ExecutedByRef string `json:"executed_by_ref"` AttestedByRef string `json:"attested_by_ref"` AttestedAt time.Time `json:"attested_at"` CollectedAt time.Time `json:"collected_at"` Devices []DeviceEvidence `json:"devices"` Sessions []SessionEvidence `json:"sessions"` } type DeviceEvidence struct { DeviceID string `json:"device_id"` HardwareModel string `json:"hardware_model"` HardwareRevision string `json:"hardware_revision"` FirmwareVersion string `json:"firmware_version"` IdentityRef string `json:"identity_ref"` AudioProfileRef string `json:"audio_profile_ref"` } type SessionEvidence struct { ConversationID string `json:"conversation_id"` DeviceID string `json:"device_id"` EngineMode string `json:"engine_mode"` Scenario string `json:"scenario"` RoomRef string `json:"room_ref"` DeviceRTCUserRef string `json:"device_rtc_user_ref"` AgentRTCUserRef string `json:"agent_rtc_user_ref"` ProviderTaskRef string `json:"provider_task_ref"` StartedAt time.Time `json:"started_at"` EndedAt time.Time `json:"ended_at"` Lifecycle []string `json:"lifecycle"` RTCJoinSucceeded bool `json:"rtc_join_succeeded"` UplinkSucceeded bool `json:"uplink_succeeded"` DownlinkSucceeded bool `json:"downlink_succeeded"` NearAudioPublished bool `json:"near_audio_published"` FarAudioPlayed bool `json:"far_audio_played"` CredentialRenewed bool `json:"credential_renewed"` NetworkRecovered bool `json:"network_recovered"` ProviderStartAccepted bool `json:"provider_start_accepted"` ProviderStopConfirmed bool `json:"provider_stop_confirmed"` FinalTranscriptReceived bool `json:"final_transcript_received"` UsageRecorded bool `json:"usage_recorded"` QualityResult string `json:"quality_result"` NoDuplicateProviderTask bool `json:"no_duplicate_provider_task"` ResourcesReleased bool `json:"resources_released"` CloudEvidenceRef string `json:"cloud_evidence_ref"` DeviceEvidenceRef string `json:"device_evidence_ref"` ProviderEvidenceRef string `json:"provider_evidence_ref"` QualityEvidenceRef string `json:"quality_evidence_ref"` UsageEvidenceRef string `json:"usage_evidence_ref"` } type Check struct { Code string `json:"code"` Status string `json:"status"` Detail string `json:"detail"` } type Report struct { SchemaVersion string `json:"schema_version"` Result string `json:"result"` GeneratedAt time.Time `json:"generated_at"` CandidateVersion string `json:"candidate_version"` BuildDigest string `json:"build_digest"` EvidenceScope string `json:"evidence_scope"` ExternalValidationRequired bool `json:"external_validation_required"` Environment string `json:"environment"` RTCProviderRef string `json:"rtc_provider_ref"` VoiceProviderRef string `json:"voice_provider_ref"` DeviceCount int `json:"device_count"` SessionCount int `json:"session_count"` CompletedEngineModes []string `json:"completed_engine_modes"` CompletedScenarios []string `json:"completed_scenarios"` ModeSessionCounts map[string]int `json:"mode_session_counts"` ScenarioSessionCounts map[string]int `json:"scenario_session_counts"` InputSHA256 map[string]string `json:"input_sha256"` Checks []Check `json:"checks"` Disclaimer string `json:"disclaimer"` } func ParseCapture(content []byte) (Capture, error) { var capture Capture if err := decodeStrict(content, &capture); err != nil { return Capture{}, fmt.Errorf("decode real device RTC capture: %w", err) } return capture, nil } func ParseReport(content []byte) (Report, error) { var report Report if err := decodeStrict(content, &report); err != nil { return Report{}, fmt.Errorf("decode real device RTC report: %w", err) } return report, nil } func Evaluate(capture Capture, candidateVersion, buildDigest string, generatedAt time.Time) Report { candidateVersion = strings.TrimSpace(candidateVersion) buildDigest = strings.ToLower(strings.TrimSpace(buildDigest)) report := Report{ SchemaVersion: ReportSchema, Result: "PASS", GeneratedAt: generatedAt.UTC(), CandidateVersion: candidateVersion, BuildDigest: buildDigest, EvidenceScope: "PROTOCOL_SIMULATION", ExternalValidationRequired: true, Environment: capture.Environment, RTCProviderRef: capture.RTCProviderRef, VoiceProviderRef: capture.VoiceProviderRef, DeviceCount: len(capture.Devices), SessionCount: len(capture.Sessions), CompletedEngineModes: []string{}, CompletedScenarios: []string{}, ModeSessionCounts: map[string]int{}, ScenarioSessionCounts: map[string]int{}, InputSHA256: map[string]string{}, Checks: []Check{}, Disclaimer: "该报告只核对同一候选的终端、RTC、语音供应商、状态、媒体、字幕、用量、质量和回收证据;协议模拟不能替代量产终端与真实供应商验收,也不等于业务发布批准。", } if capture.EvidenceSource == "HARDWARE_PROVIDER_CAPTURE" { report.EvidenceScope = "REAL_DEVICE_RTC_PROVIDER" } add := func(code string, ok bool, detail string) { status := "PASS" if !ok { status, report.Result = "FAIL", "FAIL" } report.Checks = append(report.Checks, Check{Code: code, Status: status, Detail: detail}) } add("CAPTURE_SCHEMA", capture.SchemaVersion == CaptureSchema, "RTC证据必须使用"+CaptureSchema) add("EVIDENCE_SOURCE", capture.EvidenceSource == "HARDWARE_PROVIDER_CAPTURE" || capture.EvidenceSource == "PROTOCOL_SIMULATION", "证据来源必须明确为HARDWARE_PROVIDER_CAPTURE或PROTOCOL_SIMULATION") add("CANDIDATE_IDENTITY", candidateVersion != "" && digestPattern.MatchString(buildDigest), "候选版本和sha256构建摘要必须完整") add("CAPTURE_CANDIDATE_BINDING", capture.CandidateVersion == candidateVersion && strings.ToLower(capture.BuildDigest) == buildDigest, "RTC采集必须绑定同一候选版本和构建摘要") real := capture.EvidenceSource == "HARDWARE_PROVIDER_CAPTURE" providerOK := capture.RTCProviderRef != "" && capture.VoiceProviderRef != "" && capture.RTCApplicationRef != "" && capture.RegionRef != "" environmentOK := capture.ExecutedByRef != "" && capture.AttestedByRef != "" && !capture.AttestedAt.IsZero() && !capture.CollectedAt.IsZero() && !capture.AttestedAt.After(capture.CollectedAt) if real { providerOK = providerOK && capture.RTCProviderRef != "mock" && capture.VoiceProviderRef != "mock" environmentOK = environmentOK && capture.ExecutedByRef != capture.AttestedByRef && (capture.Environment == "STAGING" || capture.Environment == "PRODUCTION") && capture.ChangeTicketRef != "" } else { environmentOK = environmentOK && capture.Environment == "LOCAL" } add("PROVIDER_IDENTITY", providerOK, "真实证据必须绑定非模拟RTC、语音供应商、应用和区域引用") add("ENVIRONMENT_IDENTITY", environmentOK, "真实证据必须绑定目标环境、变更工单、执行人和独立复核人") deviceOK := len(capture.Devices) > 0 devices := map[string]bool{} for _, device := range capture.Devices { if device.DeviceID == "" || devices[device.DeviceID] || device.HardwareModel == "" || device.HardwareRevision == "" || device.FirmwareVersion == "" || device.IdentityRef == "" || device.AudioProfileRef == "" { deviceOK = false } devices[device.DeviceID] = true } if real && len(devices) < MinimumDevices { deviceOK = false } add("DEVICE_COVERAGE", deviceOK, "真实矩阵至少覆盖3台身份、硬件、固件和音频配置完整的候选终端") sessionOK, modeOK, scenarioOK, mediaOK, lifecycleOK, recoveryOK, providerLifecycleOK, dataOK, refsOK, timelineOK := len(capture.Sessions) > 0, true, true, true, true, true, true, true, true, true seenSessions, usedDevices := map[string]bool{}, map[string]bool{} normalModes := map[string]bool{} for _, session := range capture.Sessions { if session.ConversationID == "" || seenSessions[session.ConversationID] || !devices[session.DeviceID] || session.RoomRef == "" || session.DeviceRTCUserRef == "" || session.AgentRTCUserRef == "" || session.ProviderTaskRef == "" { sessionOK = false } seenSessions[session.ConversationID] = true usedDevices[session.DeviceID] = true if !slices.Contains(RequiredEngineModes, session.EngineMode) { modeOK = false } else { report.ModeSessionCounts[session.EngineMode]++ } if !slices.Contains(RequiredScenarios, session.Scenario) { scenarioOK = false } else { report.ScenarioSessionCounts[session.Scenario]++ } if session.Scenario == "NORMAL_COMPLETION" { normalModes[session.EngineMode] = true } mediaOK = mediaOK && session.RTCJoinSucceeded && session.UplinkSucceeded && session.DownlinkSucceeded && session.NearAudioPublished && session.FarAudioPlayed lifecycleOK = lifecycleOK && slices.Equal(session.Lifecycle, RequiredLifecycle) recoveryOK = recoveryOK && (session.Scenario != "TOKEN_RENEWAL" || session.CredentialRenewed) && (session.Scenario != "NETWORK_SWITCH_RECOVERY" || session.NetworkRecovered) providerLifecycleOK = providerLifecycleOK && session.ProviderStartAccepted && session.ProviderStopConfirmed && session.NoDuplicateProviderTask && session.ResourcesReleased dataOK = dataOK && session.FinalTranscriptReceived && session.UsageRecorded && session.QualityResult == "PASS" refsOK = refsOK && session.CloudEvidenceRef != "" && session.DeviceEvidenceRef != "" && session.ProviderEvidenceRef != "" && session.QualityEvidenceRef != "" && session.UsageEvidenceRef != "" timelineOK = timelineOK && !session.StartedAt.IsZero() && !session.EndedAt.IsZero() && !session.EndedAt.Before(session.StartedAt) && !session.EndedAt.After(capture.CollectedAt) } if real { sessionOK = sessionOK && len(capture.Sessions) >= MinimumSessions && len(usedDevices) >= MinimumDevices for _, mode := range RequiredEngineModes { if report.ModeSessionCounts[mode] == 0 || !normalModes[mode] { modeOK = false } } for _, scenario := range RequiredScenarios { if report.ScenarioSessionCounts[scenario] == 0 { scenarioOK = false } } } for _, mode := range RequiredEngineModes { if report.ModeSessionCounts[mode] > 0 { report.CompletedEngineModes = append(report.CompletedEngineModes, mode) } } for _, scenario := range RequiredScenarios { if report.ScenarioSessionCounts[scenario] > 0 { report.CompletedScenarios = append(report.CompletedScenarios, scenario) } } add("SESSION_IDENTITY", sessionOK, "真实矩阵至少6通且会话、设备、房间、用户和供应商任务引用必须唯一完整") add("ENGINE_MODE_COVERAGE", modeOK, "三种语音方式均须完成正常闭环") add("SCENARIO_COVERAGE", scenarioOK, "矩阵须覆盖正常完成、凭证续签、网络切换恢复和受控停止") add("RTC_MEDIA_PATH", mediaOK, "每通必须入房成功并证明上行、下行、近端发布和远端播放") add("STATE_LIFECYCLE", lifecycleOK, "每通必须按ROOM_ISSUED到COMPLETED的受控状态序列收敛") add("TOKEN_NETWORK_RECOVERY", recoveryOK, "对应场景必须证明短期凭证续签或网络切换恢复") add("PROVIDER_LIFECYCLE", providerLifecycleOK, "供应商启停、任务去重和RTC资源回收必须完成") add("TRANSCRIPT_USAGE_QUALITY", dataOK, "每通必须形成最终字幕、用量记录和业务方质量通过结论") add("EVIDENCE_REFERENCES", refsOK, "每通必须保存云端、设备、供应商、质量和用量脱敏引用") add("EXECUTION_TIMELINE", timelineOK, "每通开始结束时间必须有效且不晚于采集时间") if real && providerOK && environmentOK && deviceOK && sessionOK && modeOK && scenarioOK && mediaOK && lifecycleOK && recoveryOK && providerLifecycleOK && dataOK && refsOK && timelineOK && report.Result == "PASS" { report.ExternalValidationRequired = false } return report } func SHA256(content []byte) string { digest := sha256.Sum256(content) return hex.EncodeToString(digest[:]) } func decodeStrict(content []byte, target any) error { decoder := json.NewDecoder(bytes.NewReader(content)) decoder.DisallowUnknownFields() if err := decoder.Decode(target); err != nil { return err } var trailing any if err := decoder.Decode(&trailing); !errors.Is(err, io.EOF) { return errors.New("trailing JSON data") } return nil }